FreeCourse Logo
FreeCourse.io
Verified CouponsFree CoursesJobsBlog
Categories
Home/Courses/Agentische KI: Von Chatbots zu Agenten, die Arbeit erledigen
Agentische KI: Von Chatbots zu Agenten, die Arbeit erledigen
IT & Software100% OFF

Agentische KI: Von Chatbots zu Agenten, die Arbeit erledigen

Udemy Instructor
0(2 students)
Self-paced
All Levels

About this course

Dieser Kurs wurde unter Zuhilfenahme von künstlicher Intelligenz erstellt. Alle verkaufen gerade „Agenten“. Manche dieser Produkte planen tatsächlich, rufen Tools auf und erledigen mehrstufige Arbeit.

Viele sind Chatbots oder feste Workflows mit neuem Etikett. Dieser Kurs gibt Ihnen einen klaren, anbieterneutralen Weg, den Unterschied zu erkennen, und führt Sie dann vom Verständnis von Agenten bis dahin, einen zu bauen, zu testen und seine Kosten zu berechnen. Er ist eine breite Einführung für Fachleute und neue Entwickler.

Sie brauchen keinen Hintergrund in maschinellem Lernen. Wenn Sie ChatGPT, Claude, Gemini oder Copilot regelmäßig nutzen und wissen wollen, was sich ändert, wenn das Modell anfängt zu handeln, ist dies der richtige Einstieg. Ein Unternehmen, ein Pilotprojekt, von Anfang bis EndeWir begleiten Harborline Logistics, eine erfundene Spedition mit einem Rückstau an E-Mails zum Sendungsstatus und verspäteten Zollpapieren.

Das Betriebsteam führt ein Agenten-Pilotprojekt über 90 Tage namens Relay durch. Jedes Konzept des Kurses wird auf dieses Pilotprojekt angewendet, sodass Sie denselben Agenten von einer ersten Schleife bis zu einem getesteten, beaufsichtigten und kalkulierten System wachsen sehen. Was der Kurs behandeltAbschnitt 1, Was agentische KI ausmacht: Chatbot, Copilot, Workflow und Agent; die Agentenschleife; fünf Autonomiestufen; wie man Agent Washing erkenntAbschnitt 2, Wie Agenten denken: Modell, Tools und Anweisungen; ReAct, Planung und Reflexion; fünf Workflow-Muster und wann man keinen Agenten nutztAbschnitt 3, Tools, MCP und Computer Use: Tool-Aufrufe, das Model Context Protocol, Browser- und Computer-Nutzung sowie Agentenmodi in ChatGPT, Claude und GeminiAbschnitt 4, Gedächtnis, Wissen und Kontext: Context Engineering, Abruf (RAG), Skills und wie man einen Agenten in Arbeitsanweisungen des Unternehmens verankertAbschnitt 5, Agenten ohne Code bauen: die No-Code-Landschaft, zuerst eine Agentenspezifikation schreiben und einen Sendungsstatus-Agenten in n8n bauenAbschnitt 6, Agenten in Python bauen: ein SDK ohne Bindung wählen, ein funktionierender Agent in etwa fünfzig Zeilen, die Anbindung an einen MCP-Server und das Lesen von TracesAbschnitt 7, Multi-Agenten-Systeme: wann ein Agent nicht reicht, Orchestratoren und Übergaben, das A2A-Protokoll und ein Zollschalter mit zwei AgentenAbschnitt 8, Vertrauen, Nutzen, Einführung: Evaluation, typische Fehlermuster und Guardrails, menschliche Freigabe, Kosten pro Aufgabe, KPIs und eine Einführungs-Roadmap über 90 TagePraxis-Labs, die auf Ihrem eigenen Computer laufenDie Bau-Labs laufen lokal mit kostenlosen Tools: Python, Ollama für lokale Modelle und n8n in Docker.

Sie brauchen keinen kostenpflichtigen API-Schlüssel. Der Starter-Ordner enthält eine Mock-API für die Sendungsverfolgung, einen MCP-Server, zwölf Arbeitsanweisungen von Harborline sowie die Test-E-Mails und Zollfälle aus den Videos. Lab 1: dieselbe Aufgabe in den Agentenmodi von ChatGPT, Claude und Gemini ausführen und ihre Aktionsprotokolle lesenLab 2: die Arbeitsanweisungen von Harborline indexieren und Antworten mit und ohne Abruf vergleichenLab 3: einen No-Code-Status-Agenten in n8n bauen und ihn mit zehn Test-E-Mails ausführenLab 4: diesen Agenten in Python nachbauen, sein Tool hinter einen MCP-Server verlegen und einen eingebauten Fehler im Trace findenLab 5: einen zweiten Agenten mit Übergabe ergänzen und Ergebnisse und Token-Verbrauch mit dem einzelnen Agenten vergleichenEin paar ehrliche Hinweise.

Lab 1 nutzt den Agentenmodus eines Chat-Assistenten in Ihrem Browser, und welche Agentenmodi Sie nutzen können, hängt von Ihrem Tarif und Ihrer Region ab. Bietet Ihrer keinen an, können Sie den aufgezeichneten Sitzungen trotzdem folgen. Lokale Modelle sind nicht deterministisch, daher weichen Ihre Antworten und Erfolgsquoten etwas von den Videos ab.

Das lokale Modell der Labs braucht etwa 6 GB freien Arbeitsspeicher. Der Python-Abschnitt ist optional, wenn Sie nur die Konzepte und den No-Code-Weg wollen. Übung und LernkontrolleJeder der acht Abschnitte hat ein Quiz und eine Praxisaufgabe mit ausgearbeiteter Musterlösung.

Im Abschlussprojekt wählen Sie eine Aufgabe aus Ihrem eigenen Job und erstellen eine Agentenspezifikation, eine Autonomiestufe mit Freigabeschritten, ein Evaluations-Set mit zehn Fällen und eine Kostenschätzung mit einer Roadmap über 90 Tage. Ein abschließender Übungstest mit 60 Fragen deckt alle acht Abschnitte ab. Was dieser Kurs nicht istEr ist kein tiefgehender Engineering-Kurs zu einem einzelnen Framework, und er verspricht weder einen Job noch ein Zertifikat noch ein bestimmtes Ergebnis.

Er ist eine Landkarte des Feldes mit genug Praxis, damit Sie Agentenvorschläge beurteilen, ein kleines Pilotprojekt durchführen und entscheiden können, was Sie als Nächstes lernen. Die Tools in diesem Feld ändern sich schnell, deshalb lehrt der Kurs portable Bausteine (Spezifikation, Prompts, Tools und Evaluationen), die bei einem Plattformwechsel erhalten bleiben.

Skills you'll gain

Other IT & SoftwareGerman

Available Coupons

Loading...

Course Information

Level: All Levels

Suitable for learners at this level

Duration: Self-paced

Total course content

Instructor: Udemy Instructor

Expert course creator

This course includes:

  • 📹Video lectures
  • 📄Downloadable resources
  • 📱Mobile & desktop access
  • 🎓Certificate of completion
  • ♾️Lifetime access
$0$89.99

Save $89.99 today!

Enroll Now - Free

Redirects to Udemy • Limited free enrollments

Share this course

https://freecourse.io/courses/agentische-ki-von-chatbots-zu-agenten-die-arbeit-erledigen

You May Also Like

Explore more courses similar to this one

IA agêntica: de chatbots a agentes que fazem o trabalho
IT & Software
0% OFF

IA agêntica: de chatbots a agentes que fazem o trabalho

Udemy Instructor

Este curso contém o uso de inteligência artificial.Todo mundo está vendendo "agentes" agora. Alguns desses produtos realmente planejam, chamam ferramentas e concluem trabalhos de vários passos. Muitos são chatbots ou fluxos fixos com um rótulo novo. Este curso dá a você uma forma clara e neutra em relação a fornecedores de perceber a diferença, e depois leva você de entender agentes a construir, testar e calcular o custo de um. É uma introdução ampla para profissionais e para quem está começando a construir. Você não precisa de formação em machine learning. Se você usa o ChatGPT, o Claude, o Gemini ou o Copilot com frequência e quer saber o que muda quando o modelo começa a agir, este é o ponto de partida certo. Uma empresa, um piloto, do início ao fimAcompanhamos a Harborline Logistics, um agente de cargas fictício com uma fila de e-mails de status de remessa e papelada aduaneira atrasada. A equipe de operações roda um piloto de agente de 90 dias chamado Relay. Cada conceito do curso é aplicado a esse piloto, então você vê o mesmo agente crescer de um primeiro loop a um sistema testado, supervisionado e com custo calculado. O que o curso cobreSeção 1, O que a IA agêntica é de fato: chatbot, copiloto, fluxo de trabalho e agente; o loop do agente; cinco níveis de autonomia; como identificar o agent washingSeção 2, Como agentes pensam: modelo, ferramentas e instruções; ReAct, planejamento e reflexão; cinco padrões de fluxo e quando não usar um agenteSeção 3, Ferramentas, MCP e uso do computador: chamada de ferramentas, o Model Context Protocol, uso do navegador e do computador, e os modos agente do ChatGPT, do Claude e do GeminiSeção 4, Memória, conhecimento e contexto: engenharia de contexto, recuperação (RAG), habilidades e a fundamentação de um agente nos procedimentos da empresaSeção 5, Agentes sem programar: o panorama no-code, escrever primeiro a especificação do agente e construir um agente de status de remessa no n8nSeção 6, Agentes em Python: escolher um SDK sem ficar preso, um agente funcional em cerca de cinquenta linhas, conectá-lo a um servidor MCP e ler tracesSeção 7, Sistemas multiagente: quando um agente não basta, orquestradores e transferências, o protocolo A2A e uma mesa aduaneira de dois agentesSeção 8, Confiança, valor e adoção: avaliação, modos de falha comuns e barreiras de proteção, aprovação humana, custo por tarefa, KPIs e um roteiro de adoção de 90 diasLaboratórios práticos que rodam no seu próprio computadorOs laboratórios de construção rodam localmente com ferramentas gratuitas: Python, Ollama para modelos locais e n8n no Docker. Nenhuma chave de API paga é necessária. A pasta inicial inclui uma API simulada de rastreamento de remessas, um servidor MCP, doze procedimentos da Harborline e os e-mails de teste e casos aduaneiros usados nos vídeos. Lab 1: rode a mesma tarefa nos modos agente do ChatGPT, do Claude e do Gemini e leia os registros de açõesLab 2: indexe os procedimentos da Harborline e compare respostas com e sem recuperaçãoLab 3: construa um agente de status no-code no n8n e rode-o com dez e-mails de testeLab 4: refaça esse agente em Python, coloque a ferramenta atrás de um servidor MCP e ache um bug plantado pelo traceLab 5: acrescente um segundo agente com uma transferência e compare resultados e uso de tokens com o agente únicoAlgumas observações honestas. O Lab 1 usa o modo agente de um assistente de chat no seu navegador, e os modos agente disponíveis dependem do seu plano e da sua região. Se o seu não oferecer um, você ainda pode acompanhar as sessões gravadas. Modelos locais não são determinísticos, então as suas respostas e taxas de acerto vão variar um pouco em relação aos vídeos. O modelo local usado nos laboratórios precisa de cerca de 6 GB de memória livre. A seção de Python é opcional se você só quiser os conceitos e o caminho no-code. Prática e avaliaçãoCada uma das oito seções tem um quiz e uma tarefa prática com uma resposta modelo comentada. O projeto final pede que você escolha uma tarefa do seu próprio trabalho e produza uma especificação de agente, um nível de autonomia com pontos de aprovação, um conjunto de avaliação de dez casos e uma estimativa de custo com um roteiro de 90 dias. Um exame prático final de 60 perguntas cobre as oito seções. O que este curso não éNão é um curso aprofundado de engenharia sobre um único framework, e não promete emprego, certificação nem um resultado específico. É um mapa da área, com prática suficiente para você avaliar propostas de agentes, rodar um piloto pequeno e decidir o que aprender em seguida. As ferramentas desta área mudam rápido, então o curso ensina peças portáteis (especificação, prompts, ferramentas e avaliações) que continuam valendo quando uma plataforma muda.

0.0•0•Self-paced
FREE$83.99
Enroll
PLC Programming: Ladder Logic for Beginners
IT & Software
0% OFF

PLC Programming: Ladder Logic for Beginners

Udemy Instructor

Learn PLC programming and ladder logic from scratch in approximately three hours—without buying expensive PLC hardware or industrial software.This beginner-friendly course takes you from understanding what a Programmable Logic Controller is to planning, building, simulating, and testing a complete automatic bottle-filling system.You will begin with the foundations of industrial automation: PLC inputs and outputs, the scan cycle, ladder diagrams, rungs, contacts, coils, bits, tags, and addresses. You will then build your first working ladder logic program using a free PLC simulator.Step by step, you will learn how to:Read and write PLC ladder logicUse Normally Open and Normally Closed contactsBuild AND, OR, XOR, and NOT logicCreate motor Start/Stop and seal-in circuitsUse standard coils and Set/Reset instructionsDetect rising and falling edges with one-shotsAvoid duplicate output and double-coil problemsControl processes with TON, TOF, and retentive timersCount products using CTU, CTD, and CTUD countersUse comparisons, range checks, math, and MOVE instructionsOrganize sequences with finite state machinesCreate clear PLC I/O lists and state diagramsUse Claude AI to draft, explain, and debug PLC logic responsiblyYou will reinforce these skills through practical exercises, including a motor Start/Stop circuit, blinking oscillator, conveyor product counter, tank-level alarm, and traffic-light state machine.For the final project, you will build an automatic bottle-filling station. The PLC program detects a bottle, fills it, allows the liquid to settle, applies a cap, ejects the finished bottle, and counts production until a complete batch is reached.The course uses free tools such as PLC Fiddle and OpenPLC, allowing you to practice PLC programming on a normal computer without physical hardware. The concepts you learn are transferable to major industrial platforms such as Siemens TIA Portal and Rockwell Studio 5000, although their software and instruction names may differ.You will also discover how an AI assistant can help create I/O lists, draft ladder logic, explain instructions, and diagnose programming problems. More importantly, you will learn the professional rule behind AI-assisted automation: understand every rung, simulate every program, test edge cases, and never treat AI-generated logic as a substitute for safety engineering.Who this course is forThis course is designed for:Complete beginners interested in PLC programmingElectrical and electronic engineering studentsAutomation and control engineering studentsMaintenance technicians entering industrial automationElectricians who want to understand ladder logicMakers and programmers exploring PLC control systemsAnyone who wants a fast, practical introduction to PLCsNo previous PLC programming experience is required. No physical PLC is required. You only need a computer, internet access, and a willingness to practice.By the end of the course, you will understand the fundamental thinking behind ladder logic and be able to build, organize, simulate, test, and explain beginner PLC automation projects

0.0•655•Self-paced
FREE$94.99
Enroll
2 Week No-Code AI Automation Certification
IT & Software
0% OFF

2 Week No-Code AI Automation Certification

Udemy Instructor

This course contains the use of artificial intelligence.The 2 Week No-Code AI Automation Certification is a practical, beginner-friendly program designed to help you build real-world AI-powered automations without writing code. In just two weeks, you will learn how to connect apps, automate repetitive tasks, use GPTs inside workflows, and design complete automation systems that save time, reduce manual work, and improve productivity.This course begins with the fundamentals of no-code AI automation, helping you understand the difference between traditional automation and intelligent automation. You will explore how businesses, creators, professionals, and teams use automation to handle emails, summarize information, update spreadsheets, create content, send notifications, and manage routine workflows. Instead of relying on manual copy-paste work, you will learn how to build systems that move data from one tool to another automatically.In Week 1, you will get hands-on with leading automation platforms like Zapier and Make. You will learn how triggers, actions, and multi-step workflows work, and when to use each platform depending on the use case. You will also learn how to connect common business tools such as Gmail, Slack, Google Sheets, forms, databases, and APIs without needing programming experience. The course introduces simple but powerful concepts like data flow, inputs, outputs, mapping fields, and connecting multiple apps together.A major focus of this certification is using AI inside automations. You will learn how to use GPT-based tools to generate responses, summarize long messages, transform raw data, classify information, draft content, and personalize communication. By the end of Week 1, you will complete your first working automation and understand how to think like an automation designer.In Week 2, you will move into real-world automation builds. You will create workflows for email automation, communication follow-ups, content creation pipelines, reporting dashboards, and data processing systems. You will also learn how to build conditional automations using branching logic, filters, paths, and basic error handling. These skills are essential for creating workflows that are reliable, scalable, and useful in real business environments.By the end of this course, you will be able to design, build, test, and optimize production-ready no-code AI automation systems. You will understand how to identify automation opportunities, choose the right tools, connect apps, add AI intelligence, and create workflows that improve efficiency. Whether you are a business owner, freelancer, marketer, educator, operations professional, or complete beginner, this course will help you turn everyday tasks into smart automated systems using no-code tools, AI workflows, and practical automation thinking.

4.6•1.4K•Self-paced
FREE$97.99
Enroll
FreeCourse LogoFreeCourse

Freecourse.io brings you high-quality online courses with free certificates to help you upskill, boost your career, and achieve your goals anytime, anywhere.

Resources

  • Courses
  • Jobs
  • Categories
  • Features

Company

  • About
  • Blog
  • Contact

Legal

  • Privacy
  • Terms
  • Cookies
  • Licenses

© 2026 FreeCourse. All rights reserved.